I advocate Smith-IRV because it is simple and meets Condorcet
Winner/Loser and Smith, as an improved IRV it sidesteps FairVote's
attempts to redefine RCV as the Hare Method. Against more complex
Condorcet Methods it can be held as Later No Harm/Help Sufficient when
no Condorcet Winner is present (by which I mean the possible violations
can be argued to be small enough to ignore). The simplicity and
demonstrably lower LNH/H in my view outweigh any advantages the more
complex Condorcet methods have.
The BTR-IRV in this discussion is an interesting alternative. It
requires a maximum of N-1 pairings, making it more handcountable. like
the acronym BTR-IRV, because it suggests Better IRV. In the absence of a
Condorcet Winner it would have a greater Later Harm/Help effect than
Smith-IRV, because each pairing is an opportunity for a Later Harm
effect, where after the Smith Set is found Smith-IRV is LNH/H safe.
Both methods meet Condorcet's criteria, BTR-IRV is a little simpler and
more hand count friendly, Smith IRV meets the additional Smith Criteria
and has less LNH/H potential. This also presents more of a spectrum. If
the biggest concern is LNH/H, then choose Hare, since Hare has
weaknesses (the cost of LNH compliance), there are two Better IRV
methods, Smith and BTR, Smith has better LNH than BTR, but BTR is
simpler and can be hand counted.
It won't take much work to implement BTR-IRV in Vote::Count , but right
now I have a lot of Work in Progress for STV methods, so may not get to